Artificial Intelligence (AI), particularly in the form of intelligent AI agents, is transforming education, industry, and everyday life. These agents extend the capabilities of Large Language Models (LLMs) by integrating planning, decision-making, tool use, and multi-agent collaboration, enabling systems that can reason, adapt, and act in dynamic environments. As such systems become integral to modern workplaces and everyday problem solving, early exposure equips high school students with systems thinking, practical problem-solving skills, and ethical awareness, while preparing them to create applications that address real-world needs. Yet most high school AI programs focus on basic model usage and overlook the skills required to design and deploy agentic systems. Existing resources are largely aimed at university learners and assume substantial programming expertise, creating a significant accessibility gap. To address this need, we present a structured hackathon-based framework for introducing high school students to the design and application of AI agents. The framework combines expert-led lectures on core topics such as agent architectures, prompting strategies, reasoning methods, and tool-use protocols with a guided hackathon in which students collaboratively develop domain-specific agent-based chatbots. We provide a complete suite of instructional materials, step-by-step tutorials, and starter code to support hands-on learning, enabling participants to build functional agents capable of reasoning and interacting with external tools. Our approach bridges the gap between AI literacy and practical deployment while fostering creativity, collaboration, and responsible innovation, and our findings suggest that early engagement with agent-based AI design equips students with both technical proficiency and the mindset to shape the AI-driven future.

The softmax-contaminated mixture of experts (MoE) model is deployed when a large-scale pre-trained model, which plays the role of a fixed expert, is fine-tuned for learning downstream tasks by including a new contamination part, or prompt, functioning as a new, trainable expert. Despite its popularity and relevance, the theoretical properties of the softmax-contaminated MoE have remained unexplored in the literature. In the paper, we study the convergence rates of the maximum likelihood estimator of gating and prompt parameters in order to gain insights into the statistical properties and potential challenges of fine-tuning with a new prompt. We find that the estimability of these parameters is compromised when the prompt acquires overlapping knowledge with the pre-trained model, in the sense that we make precise by formulating a novel analytic notion of distinguishability. Under distinguishability of the pre-trained and prompt models, we derive minimax optimal estimation rates for all the gating and prompt parameters. By contrast, when the distinguishability condition is violated, these estimation rates become significantly slower due to their dependence on the prompt convergence rate to the pre-trained model. Finally, we empirically corroborate our theoretical findings through several numerical experiments.

Generalized sliced-Wasserstein distance is a variant of slicedWasserstein distance that exploits the power of non-linear projection through a given defining function to better capture the complex structures of probability distributions. Similar to the sliced-Wasserstein distance, generalized slicedWasserstein is defined as an expectation over random projections which can be approximated by the Monte Carlo method. However, the complexity of that approximation can be expensive in high-dimensional settings. To that end, we propose to form deterministic and fast approximations of the generalized sliced-Wasserstein distance by using the concentration of random projections when the defining functions are polynomial function and neural network type function. Our approximations hinge upon an important result that one-dimensional projections of a high-dimensional random vector are approximately Gaussian.

Pruning at initialization (PaI) aims to remove weights of neural networks before training in pursuit of training efficiency besides the inference. While off-the-shelf PaI methods manage to find trainable subnetworks that outperform random pruning, their performance in terms of both accuracy and computational reduction is far from satisfactory compared to post-training pruning and the understanding of PaI is missing. For instance, recent studies show that existing PaI methods only able to find good layerwise sparsities not weights, as the discovered subnetworks are surprisingly resilient against layerwise random mask shuffling and weight re-initialization. In this paper, we study PaI from a brand-new perspective -- the topology of subnetworks. In particular, we propose a principled framework for analyzing the performance of Pruning and Initialization (PaI) methods with two quantities, namely, the number of effective paths and effective nodes. These quantities allow for a more comprehensive understanding of PaI methods, giving us an accurate assessment of different subnetworks at initialization. We systematically analyze the behavior of various PaI methods through our framework and observe a guiding principle for constructing effective subnetworks: *at a specific sparsity, the top-performing subnetwork always presents a good balance between the number of effective nodes and the number of effective paths.* Inspired by this observation, we present a novel data-agnostic pruning method by solving a multi-objective optimization problem. By conducting extensive experiments across different architectures and datasets, our results demonstrate that our approach outperforms state-of-the-art PaI methods while it is able to discover subnetworks that have much lower inference FLOPs (up to 3.4$\times$). Code will be fully released.